Explain Vector Database vs Inverted Index.

A vector database retrieves results based on semantic similarity using embeddings, while an inverted index matches exact terms to documents for precise keyword search.

When to Use

• Vector Database: semantic search, recommendations, deduplication, anomaly detection, Retrieval-Augmented Generation (RAG).
• Inverted Index: keyword queries, faceted filters, logs search, compliance audits, prefix matches.

Example

Search: “How to improve laptop battery life”

• Vector DB: returns “optimize power settings” (semantic match).
• Inverted Index: returns documents with “battery life” keywords.

Want to master these concepts?

Explore:

• Grokking System Design Fundamentals
• Grokking the Coding Interview
• Mock Interviews with ex-FAANG engineers

These resources will help you connect theory with real interview practice.

Why Is It Important

Hybrid search systems often combine both to maximize relevance, recall, and precision—a key interview and system design skill.

Interview Tips

• Start by defining both clearly.
• Contrast retrieval: cosine similarity/ANN vs term-to-document mappings (BM25).
• Mention hybrid search, RAG, and real-world trade-offs.

Trade-offs

• Vector DB: richer recall, but higher cost, memory, and latency.
• Inverted Index: fast and cheap, but weak at handling synonyms, typos, or paraphrases.

Pitfalls

• Assuming vectors fully replace keywords.
• Ignoring embedding drift/versioning.
• Overlooking evaluation metrics (nDCG, Recall@k).
• Storing embeddings without compression or HNSW/IVF tuning.